class: center, middle, inverse, title-slide # Presentation: ## subtitle ### Stu Field ### Thursday December 09, 2021 --- <!-- class: title-slide, left, top --> <!-- # Presentation:<br> subtitle --> <!-- ### Stu Field --> <!-- #### 2021-12-09 --> <!-- --- --> # Outline - Something - This is the title: Presentation: - This is the author: Stu Field - That other thing -- - on mouse click -- > Here is some code that appears as a quote -- - And we are done --- layout: false class: inverse, middle, center # Heading Text --- # Some Acronyms and Codes .pull-left[ Test or code chunks for the lhs ] .pull-right[ rhs bits ] --- # Plot on the entire slide with a hex sticker <img src="hex_images/tidyposterior.png" class="title-hex"> <img src="hex_images/purrr.png" class="title-hex"> ```r ggplot(mtcars, aes(x = disp, y = mpg, color = factor(cyl))) + geom_point(size = 5) ``` <img src="data:image/png;base64,#template_files/figure-html/example-plot-1.svg" width="40%" style="display: block; margin: auto;" /> --- # Plots on one side; Left side on Click .pull-left[ show some code here but don't make image: ```r ggplot(mtcars, aes(x = disp, y = mpg, color = factor(cyl))) + geom_point(size = 5) ``` ] -- .pull-right[ <img src="data:image/png;base64,#template_files/figure-html/example-plot-again-1.svg" width="90%" style="display: block; margin: auto;" /> ] --- # Insert images from file <img src="hex_images/dplyr.png" class="title-hex"> .pull-left[ Most statisticians say "no" <img src="data:image/png;base64,#figures/nopestradomis.jpg" width="450px" style="display: block; margin: auto;" /> ] .pull-right[ The Variance-Bias Trade-off <img src="data:image/png;base64,#figures/VarBias.png" width="450px" style="display: block; margin: auto;" /> ] --- # The Variance-Bias Trade-off <img src="hex_images/rstudio.png" class="title-hex"> .pull-left[ * Variance: Differences in model output performance when refitting slightly different data - fits idiosyncrasies (noise) rather than signal - model is too flexible - overfitting ] -- .pull-right[ * Bias: lack of model performance due to model missing relationships between response and predictors - doesn't fit enough signal in the data - model is too rigid - underfitting ] --- # The Packages of the tidyverse <img src="hex_images/tibble.png" class="title-hex"> <img src="hex_images/forcats.png" class="title-hex"> <img src="data:image/png;base64,#hex_images/tidyverse_all.png" width="45%" style="display: block; margin: auto;" /> --- # Flow-Diagram of the Cross-validation process <img src="hex_images/tidyverse.png" class="title-hex"> <img src="data:image/png;base64,#figures/crossval-tree.png" width="75%" style="display: block; margin: auto;" />